Microbial dynamics during harmful dinoflagellate Ostreopsis cf. ovata growth: Bacterial succession and viral abundance pattern

Abstract Algal–bacterial interactions play a major role in shaping diversity of algal associated bacterial communities. Temporal variation in bacterial phylogenetic composition reflects changes of these complex interactions which occur during the algal growth cycle as well as throughout the lifetime of algal blooms. Viruses are also known to cause shifts in bacterial community diversity which could affect algal bloom phases. This study investigated on changes of bacterial and viral abundances, bacterial physiological status, and on bacterial successional pattern associated with the harmful benthic dinoflagellate Ostreopsis cf. ovata in batch cultures over the algal growth cycle. Bacterial community phylogenetic structure was assessed by 16S rRNA gene ION torrent sequencing. A comparison between bacterial community retrieved in cultures and that one co‐occurring in situ during the development of the O. cf. ovata bloom from where the algal strain was isolated was also reported. Bacterial community growth was characterized by a biphasic pattern with the highest contributions (~60%) of highly active bacteria found at the two bacterial exponential growth steps. An alphaproteobacterial consortium composed by the Rhodobacteraceae Dinoroseobacter (22.2%–35.4%) and Roseovarius (5.7%–18.3%), together with Oceanicaulis (14.2‐40.3%), was strongly associated with O. cf. ovata over the algal growth. The Rhodobacteraceae members encompassed phylotypes with an assessed mutualistic‐pathogenic bimodal behavior. Fabibacter (0.7%–25.2%), Labrenzia (5.6%–24.3%), and Dietzia (0.04%–1.7%) were relevant at the stationary phase. Overall, the successional pattern and the metabolic and functional traits of the bacterial community retrieved in culture mirror those ones underpinning O. cf. ovata bloom dynamics in field. Viral abundances increased synoptically with bacterial abundances during the first bacterial exponential growth step while being stationary during the second step. Microbial trends also suggest that viruses induced some shifts in bacterial community composition.

shifts in bacterial community diversity which could affect algal bloom phases. This study investigated on changes of bacterial and viral abundances, bacterial physiological status, and on bacterial successional pattern associated with the harmful benthic dinoflagellate Ostreopsis cf. ovata in batch cultures over the algal growth cycle.
Bacterial community phylogenetic structure was assessed by 16S rRNA gene ION torrent sequencing. A comparison between bacterial community retrieved in cultures and that one co-occurring in situ during the development of the O. cf. ovata bloom from where the algal strain was isolated was also reported. Bacterial community growth was characterized by a biphasic pattern with the highest contributions (~60%) of highly active bacteria found at the two bacterial exponential growth steps. An alphaproteobacterial consortium composed by the Rhodobacteraceae Dinoroseobacter (22.2%-35.4%) and Roseovarius (5.7%-18.3%), together with Oceanicaulis (14.2-40.3%), was strongly associated with O. cf. ovata over the algal growth. The Rhodobacteraceae members encompassed phylotypes with an assessed mutualisticpathogenic bimodal behavior. Fabibacter (0.7%-25.2%), Labrenzia (5.6%-24.3%), and Dietzia (0.04%-1.7%) were relevant at the stationary phase. Overall, the successional pattern and the metabolic and functional traits of the bacterial community retrieved in culture mirror those ones underpinning O. cf. ovata bloom dynamics in field. Viral abundances increased synoptically with bacterial abundances during the first bacterial exponential growth step while being stationary during the second step. Microbial trends also suggest that viruses induced some shifts in bacterial community composition.
A deep knowledge on phylogenetic composition and successional dynamics of bacterial communities associated with HABs is therefore recognized as a crucial step for unveiling relevant and recurrent algal-bacterial associations underpinning the different bloom phases (Bagatini et al., 2014;Mayali et al., 2011;Tada, Taniguchi, Sato-Takabe, & Hamasaki, 2012;Yang et al., 2015), and in parallel with complementing laboratory-based studies, it will allow to elucidate the functional significance of these complex interactions (Bagatini et al., 2014;Buchan et al., 2014;Kazamia, Helliwell, Purton, & Smith, 2016;Sison-Mangus et al., 2014). Indeed, although 16S rRNA gene phylogenetic surveys do not directly decode bacterial functionality, they still provide insights on how the different bacterial groups correlate within the assemblages and with the microalgal partner, considering certain metabolic characteristics significant to the groups and to the associated organism (Amin et al., 2012;Buchan et al., 2014;Gifford, Sharma, & Moran, 2014;Newton et al., 2010). Next Generation Sequencing approaches typically allow a deeper phylogenetic analysis than traditional molecular methods, used in most of the available studies describing bacterial communities associated with toxic dinoflagellates (Garcés et al., 2007;Jones et al., 2010;Mayali et al., 2011;Park et al., 2015;Yang, Zhou, Zheng, Tian, & Zheng, 2012), therefore considerably reducing the gap of knowledge on this topic.
Additionally, the highly respiring bacteria were identified as those ones able to reduce the fluorogenic redox dye 5-cyano-2,3-dytolyl tetrazolium chloride (CTC), in order to provide details on bacterial community's physiological status during cultures progression.
The presence of viruses and their abundance pattern were also evaluated synoptically throughout the O. cf. ovata growth. Actually, it is known that bacterial communities are also shaped in terms of diversity and dynamics by viral activity, mainly affecting the most abundant and metabolically active species (Del Giorgio & Gasol, 2008;Fuhrman, 1999;Sime-Ngando, 2014;Wommack & Colwell, 2000 and references therein). However, viruses have been seldom taken into consideration in HABs dynamics (Loureiro, Reñé, Garcés, Camp, & Vaqué, 2011;Meyer et al., 2014).
To the best of our knowledge, this is the first study that provides viral and highly respiring bacterial cells (CTC + cells) abundance trends, as well as bacterial 16S rRNA gene Next Generation Sequencing data associated with a cultured toxic dinoflagellate.

| Experimental setup and culture conditions
O. cf. ovata strain OOAP1209 was isolated in September 2012 from macrophyte samples collected at the early phase of an O. cf. ovata bloom along the coast of North-western Adriatic Sea (Passetto, Italy, 43°36′38″ N and 13°32′20″ E; Vanucci, Guidi, et al., 2016), using capillary pipette method under sterile conditions (Hoshaw & Rosowski, 1973), and using 0.22μm-pore-size filtered and autoclaved seawater for cell washing steps. After initial growth in microplates, cells were maintained in sterile flasks sealed with cotton plugs at 20°C ± 1°C under a 16:8 hr light:dark cycle in a growth chamber (photon flux density 110-120 μmol m 2 s −1 by cool white lamp).
Cultures were set up in sterile f/2 medium (minus silicate) (Guillard, 1975) plus selenium, with macronutrients (NO 3ˉ and PO 4 3ˉ) added at a fivefold diluted concentration. The medium was prepared from natural seawater kept several weeks in the dark before use. The seawater was 0.22-μm-pore-size filtered and autoclaved, and adjusted to salinity value of 36. Medium also contains trace metals and vitamins (Guillard, 1975). Aliquots for O. cf. ovata enumeration, bacterial and viral enumeration, and for assessment of bacterial physiological status were collected at day 0, 3, 6, 9, 12, 18, 24, 32, and 42. Aliquots for nutrient analysis were collected at day 0, 3, 6, 9, 12, 24, and 42, whereas aliquots for phylogenetic analysis of the bacterial community were collected at day 0, 6, 24, and 42. For all the analyses, aliquots were collected from each flask in triplicate.

| Ostreopsis cf. ovata enumeration and nutrient analysis
O. cf. ovata cell counts were carried out following Utermöhl method (Hasle, 1978) using a Zeiss Axiovert 100 inverted microscope at 320× magnification under bright field and phase contrast illumination. Specific growth rate (μ, day −1 ) was calculated using the following equation: where N 0 and N 1 were cell density values (cells mL −1 ) at time t 0 and t 1 .
Nitrate and phosphate analyses were performed on filtered culture medium aliquots (Whatman GF/F filters, pore size 0.7 μm) and analyzed spectrophotometrically (UV/VIS, JASCO 7800, Tokyo, Japan) according to Strickland and Parsons (1972

| Bacterial and viral enumeration and assessment of bacterial physiological status
Bacterial and virus like particles (VLPs) abundances were determined in the same culture aliquots fixed with 0.02 μm prefiltered formaldehyde (2%), following the method described by Shibata et al. (2006).
Bacterial physiological status was assessed by determining highly respiring bacteria as those able to reduce 5-cyano-2,3-ditolyl tetrazolium chloride (CTC; Sigma-Aldrich), which turns into a red fluorescent formazan detectable by epifluorescence microscopy (Sherr, del Giorgio, & Sherr, 1999). Sample aliquots (0.9 mL) were amended with 100 μL of a 50 mmol L −1 CTC solution (final concentration 5 mmol L −1 ) immediately following collection and were incubated for 3 hr in the dark at room temperature. After the incubation, samples were fixed with 0.22 μm prefiltered formaldehyde (2%) and then filtered onto 0.22 μm pore size black-stained polycarbonate membrane filters (Millipore). Cell counts were performed using epifluorescence microscopy as described above for bacteria and VLPs. After sequencing, the individual sequence reads were filtered by the PGM software for low quality and polyclonal sequences removal.

| Bacterial DNA extraction and PCR amplification
Sequences matching the PGM 3′ adaptor were also automatically trimmed. All PGM quality approved, trimmed and filtered data were exported as fastq files.

| Sequence processing and diversity analysis
The fastq files were processed using MOTHUR (Schloss et al., 2009 Sequences were submitted to GenBank with the project reference (BioProject ID) PRJNA339161.

| Statistical analysis
All statistical analyses were performed with PAST 3.14. Differences in the investigated variables were tested by the analysis of variance (ANOVA). Statistical significance was set at p < .05 for all the analyses.

| Ostreopsis cf. ovata cell growth, bacterial abundance and physiological status, viral abundance
Growth curve of O. cf. ovata is shown in Figure 1. Cultures initial cell densities were 372 ± 37 cells mL −1 ; the exponential phase ended by day 9 (mean growth rate: 0.22 ± 0.01 day −1 ) attaining to a cell yield of 2.63 × 10 3 ± 9.55 × 10 1 cells mL −1 , and at the stationary phase an increase in mucilaginous cell aggregates was evident.
Inorganic nutrients (i.e., NO 3ˉ and PO 4 3ˉ) were rapidly taken up by the cells during the first days of growth, and being almost depleted by day 12 ( Figure S1).
Over the O. cf. ovata growth cycle bacterial cell densities increased by more than one order of magnitude (range: 7.24 × 10 5 to 2.01 × 10 7 cells mL −1 , day 0 and 24, respectively; mean value: 9.41 × 10 6 ± 6.77 × 10 6 cells mL −1 ). Bacterial community growth was characterized by a biphasic pattern ( Figure 1): a first exponential phase (first bacterial growth step) occurred synoptically with the algal exponential growth phase (i.e., days 0-9), whereas a second exponential phase (second bacterial growth step) occurred between day 12 and day 24 of the algal mid stationary phase (days 9-24), and it was characterized by a lower bacterial growth rate with respect to the first one (μ = 0.24 and 0.10 day −1 , days 0-9 and 12-24, respectively; ANOVA, p < .01).
Particularly, the highest CTC + cells relative abundances were found synoptically with the first and the beginning of the second bacterial exponential steps (62.8% and 59.9%, day 6 and 12, respectively), whereas a significant drop in CTC + cells contribution was recorded at Abundance of virus like particles (VLPs) ranged between 1.29 × 10 7 and 5.50 × 10 7 VLPs mL −1 (day 0 and 9, respectively), showing a fourfold higher mean value (3.86 × 10 7 ± 1.33 × 10 7 VLPs mL −1 ) than bacterial one. While during the first bacterial growth step (days 0-9) viral abundances exhibited a synoptic increasing pattern, during the second bacterial growth step and afterward they were almost stationary, slightly decreasing (Figure 1). The consequent mean virus to bacteria ratios (VBR, Figure S2) were equal to 11.4 between days 0-9, to 4.9 between days 12-24 and to 2.3 between days 32-42.

| Microbial dynamics during Ostreopsis cf. ovata growth
Growth pattern of O. cf. ovata and nutrients temporal trend observed in this study are consistent with those previously described for the same algal species under comparable culture conditions (Pezzolesi et al., 2014. Bacterial community growth showed a biphasic pattern, characterized by two exponential growth steps having different growth rates that appear mainly triggered by different quality and amount of available substrate. The first and faster growth step, occurring synoptically with the algal exponential growth phase, suggests a rapid utilization of the available inorganic nutrients present in the medium not only by O. cf. ovata, but also by bacteria along with photosynthetic products, mostly of low molecular weight (Buchan et al., 2014;Wagner-Döbler et al., 2010).
Whereas, the second and slower bacterial growth step, occurring at the algal mid stationary phase, suggests the proliferation of bacteria able to grow on a wider pool of algal-derived organic matter including high molecular weight compounds (Buchan et al., 2014;Thornton, 2014) under low inorganic nutrients concentrations in the culture medium. Accordingly, the highest contributions of highly respiring bacteria (CTC + cells) to the community (~60% of the total bacterial cells) T A B L E 1 Bacterial diversity parameters during the Ostreopsis cf. ovata growth. Summary of total sequences after normalization (Reads), richness as number of bacterial operational taxonomic units detected at 97% identity (OTUs), Shannon diversity (H'), and Good's sample coverage obtained by Ion torrent sequencing data at the time of inoculum (day 0) and during the different algal growth phases (days 6, 24, and 42). where the rates of leucine and thymidine uptake increased and then decreased in line with the initiation and maintenance bloom phases, respectively (Meyer et al., 2014). In addition, temporal microbial (CTC + bacteria and viruses) patterns and the decreasing trend of the mean virus to bacteria ratio (VBR: from 11.4 to 2.3, days 0-9 and 32-42, respectively) indicate a more relevant viral top-down control (e.g., Meyer et al., 2014;Wommack & Colwell, 2000) at the first bacterial growth step than afterward, suggesting that viruses likely affected the bacterial community composition by impacting most active bacteria rather than affecting the alga straightly, in accordance with previous reports on bloom dynamics of Karenia brevis (Meyer et al., 2014;Paul et al., 2002). Nevertheless, further studies are needed to assess a possible presence of algal viruses (and their forms of infection) and its relative importance in the microbial dynamics. Bacterial and viral temporal patterns in this study also suggest a more tight relationship between viral abundance and bacterial growth rate rather than between viral and bacterial abundances, as similarly found in natural environments (e.g., Corinaldesi et al., 2003;Danovaro, Corinaldesi, Filippini, Fischer, & Gessner, 2008;Danovaro et al., 2011;Del Giorgio & Gasol, 2008;Middelboe, 2000;Sime-Ngando, 2014). Consistently, a higher relative abundance of fast-growing bacteria was retrieved at the first growth step than at the second one and afterward (i.e., Alphaproteobacteria with respect to Sphingobacteria, Figure 3). 1621-2214 and 5.28-6.36, OTU richness and Shannon diversity, respectively; Vanucci, Guidi, et al., 2016). This finding remarks that algal cells isolation procedure and laboratory maintenance over successive subcultures can reduce bacterial diversity of the community co-occurring with the alga in the natural environment. In fact, it is F I G U R E 4 Relative contribution of the major bacterial genera (≥1% in at least one of the samples) retrieved in O. cf. ovata batch cultures at the time of inoculum (day 0) and during the different algal growth phases (days 6, 24, and 42), as revealed from ION torrent sequencing data. "others" represent the genera with less than 1% of relative abundance individually
Overall, in this study, the co-dominance of Oceanicaulis and Dinoroseobacter phylotypes (closely related to Oceanicaulis alexandrii and Dinoroseobacter shibae at 96% and 99% 16S rRNA gene sequence similarity, respectively; Table 3) at exponential and late stationary algal growth phases reflects their high metabolic plasticity, considering the deep differences in terms of inorganic nutrient concentrations and organic matter quality and availability between the two distinct phases. in Oceanicaulis phylotypes (Oh et al., 2011). Additionally, the versatile chemoheterotrophic metabolism reported for this genus (Chen, Sheu, Chen, Wang, & Chen, 2012;Oh et al., 2011;Strompl, 2003) also encompasses efficient phosphate uptake capacity in carbonlimited medium and inorganic nutrient depleted conditions through high-affinity phosphate transporters located in the prosthecae (McAdams, 2006;Oh et al., 2011). Dinoroseobacter shibae strains have been firstly retrieved in association with toxic cultured benthic and T A B L E 2 Relative contribution (%) of the abundant OTUs (≥1% of the total reads in at least one of the samples) retrieved at the time of inoculum (day 0) and during the different algal growth phases (days 6, 24, and 42), as revealed by ION torrent sequencing data  -Döbler et al., 2010). High contributions of this species (>22%, Figure 4) were observed here all along the O. cf. ovata growth. A mutualistic-pathogenic bimodal behavior in response to algal physiological status has been demonstrated for D. shibae in co-culture with toxic dinoflagellates. Specifically, the bacterium is able to switch from a mutualistic phase, when it synthesizes vitamins B 1 and B 12  and antibacterial compounds primarily in exchange for the algal-released dimethylsulfoniopropionate, to a pathogenic phase triggered by algal senescence signaling molecules (Wagner-Döbler et al., 2010;Wang et al., 2014Wang et al., , 2015. A similar behavior has been found for other related Rhodobacteraceae (i.e., Phaeobacter gallaeciensis, P. inhibens, Seyedsayamdost et al., 2011;Segev et al., 2016;Wang et al., 2016;Rugeria pomeroyi, Riclea et al., 2012;Silicibacter sp., Sule & Belas, 2013). O. cf. ovata produces dimethylsulfoniopropionate , whereas its vitamin requirements are still unknown. A focus on O. cf. ovata B vitamins demand and on the dinoflagellate's potential vitamin uptake through its associated bacterial community warrants future research, considering that many harmful dinoflagellates show auxotrophy for some B vitamins (Croft, Lawrence, Raux-Deery, Warren, & Smith, 2005;Cruz-López & Maske, 2016;Koch et al., 2014;Tang, Koch, & Gobler, 2010). In this study, Roseovarius accounted for almost 20% of the total bacteria at O. cf. ovata exponential growth phase. Roseovarius-affiliated phylotypes have been recovered from different cultured marine algal species Onda et al., 2015), also concurrently with Oceanicaulis (Abby et al., 2014;Kuo & Lin, 2013) and Fabibacter relatives (Green, Echavarri-Bravo, Brennan, & Hart, 2015), and in association with toxic dinoflagellate blooms (Vanucci, Guidi, et al., 2016;Yang et al., 2015). Metagenomic and biochemical analyses highlighted the large metabolic portfolio of Roseovarius (Bruns et al., 2013;Riedel et al., 2015), including synthesis of dual nature compounds (i.e., algal growth promoting and algicidal ones; Ziesche et al., 2015). However, Roseovarius as well as Labrenzia strains have been shown to require both vitamin B 1 and B 7 for the growth (Biebl, Lu, Schulz, Allgaier, & Wagner-Döbler, 2007;. The comparison between laboratory and environmental data reveals that the alphaproteobacterial consortium retrieved in O. cf. ovata cultures was phylogenetically closely related to that one found during the O. cf. ovata bloom, the latter composed by the Rhodobacteraceae Ruegeria, Jannaschia, and Roseovarius together with Erythrobacter (Vanucci, Guidi, et al., 2016). Besides, the members forming the two consortia altogether share comparable metabolic traits, including species-specific de novo B vitamins synthesis and a bimodal behavior with the ability to synthesize both antibacterial and algicidal compounds (Newton et al., 2010;Pujalte, Lucena, Ruvira, Arahal, & Macián, 2014;Ziesche et al., 2015), suggesting some degree of functional similarity and redundancy. In fact, Jannaschia and Ruegeria phylotypes were still present in cultures, although in lower abundances (Table S3). Thus, while culture conditions partially modify the relative importance of lower-order taxa composing the environmental bacterial community, the overall metabolic and functional profile seems someway maintained.
Inhibitory/algicidal activity has been also strongly suggested for  (Bagatini et al., 2014;Basu, Deobagkar, Matondkar, & Furtado, 2013;Vanucci, Guidi, et al., 2016). During the O. cf. ovata bloom, however, Ilumatobacter phylotypes were the main representatives of Actinobacteria at maintenance/decline phase (Vanucci, Guidi, et al., 2016), as found in diatoms degradation processes (Bagatini et al., 2014;Zakharova et al., 2013). This finding suggests a more intimate relationship between Ilumatobacter and diatoms co-occurring at the bloom rather than with O. cf. ovata. Whereas, consistently with the results reported here, Dietzia relatives have been isolated during the termination of the planktonic harmful dinoflagellate Cochlodinium polykrikoides blooms (Kim et al., 2008), suggesting a more recurrent interaction of this bacteria with the dinoflagellates.

| CONCLUSIONS
In this study, an alphaproteobacterial consortium composed by the In order to gain insight into the functional significance and metabolic exchanges underpinning these complex interactions, future experimentation is required in defined co-cultures based on O. cf. ovata and bacterial isolates selected among those composing the algal associated community retrieved in this study. A focus on the bacterial phylotypes with an assessed mutualistic-pathogenic bimodal behavior, in response to algal physiological status, which could have relevance in O. cf. ovata bloom initiation and termination phases, it is suggested.
With respect to viral lytic activity, bacterial abundance pattern and bacterial successional trend found in this study suggest investigation on viral host specificity for the most abundant Alphaproteobacteria associated with O. cf. ovata, particularly at the first bacterial growth step. An exception can be made with regard to D. shibae, known to harbor the most complex Rhodobacteraceae' viral defense system retrieved to date (Wagner-Döbler et al., 2010).
At the same time, the presence of viruses specific for O. cf. ovata and their forms of infection (Sime-Ngando, 2014) should be also investigated.

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.

ETHICAL APPROVAL
This article does not contain any studies with human participants or animals performed by any of the authors.